Terrakotta Started in Email Deliverability and Ended Up Building an AI Dialer

The Macro: Cold Email Is Dying and Everyone Knows It

The cold email infrastructure market has been in a slow-motion crisis for about two years now. Inbox providers keep tightening filters. Sender reputation is harder to build and easier to destroy. The tools that promised “guaranteed inbox placement” are fighting a losing battle against detection systems that get smarter every quarter.

The numbers tell the story. Average cold email open rates have dropped from the mid-40s to the low-20s for most B2B senders, according to industry benchmarks from Mailchimp and HubSpot. Reply rates on cold outbound sequences have cratered even harder. SDR teams that built their entire pipeline on high-volume email are scrambling for alternatives.

This has created an interesting market moment. On one side, you have email deliverability tools like Instantly, Smartlead, and Warmbox that are doubling down on inbox warming, domain rotation, and sender reputation management. They’re treating the problem as primarily technical: if you can just get the email delivered, the rest works itself out. On the other side, you have a growing camp that thinks the channel itself is degrading and the answer is to go multi-channel or move to voice entirely.

The pivot from “fix email deliverability” to “AI-powered phone outreach” is happening across several companies simultaneously, which tells you it’s probably a real trend and not just one team’s idiosyncratic choice.

Sales tech spending remains robust. Gartner estimates the CRM and sales engagement market at over $80 billion, and the slice going to AI-augmented outbound tools is growing faster than the category average. The appetite for anything that can fill pipeline without hiring more SDRs is basically bottomless at mid-market companies.

The Micro: Three Founders, Two Pivots, One Niche

Terrakotta came through Y Combinator’s Winter 2024 batch with an email deliverability pitch. The founding team brings an unusual mix: Vincent Wilson, the CEO, was a CS student at University of Washington with an incoming offer from Palantir before he founded the company. Kusiima Boswell, the COO, came from McKinsey as a Senior Analyst after graduating from Wharton and Penn Engineering. Gabriel Wilson built StoreStash, a gig economy platform, and worked as a software engineer at Lyft before joining as the third founder.

The original product helped SDR teams diagnose and fix email reputation issues. That’s a real problem, and there are real customers who pay for it. But at some point, the team appears to have made a strategic decision to go deeper into a specific vertical rather than staying horizontal.

The current version of Terrakotta is an AI-powered prospecting platform built specifically for commercial real estate brokerages. The feature set includes automated dialing, AI cold calling capabilities, a commercial real estate GPT, and a brokerage data solution. The voice cloning component is particularly worth noting: the system can generate personalized AI voicemails using cloned voices, which is the kind of feature that sits right on the line between clever and unsettling.

Commercial real estate is actually a smart vertical choice for this kind of tool. CRE brokers are high-volume prospectors by nature. They make dozens of calls a day. They work from lists. They have standardized pitches. And they’re generally not early adopters of new software, which means the competition is weaker than you’d expect. Most CRE firms are still using basic CRM systems and manual dialing.

The website when I visited it was more application shell than marketing site, with Intercom chat widgets and AWS monitoring scripts visible in the page source. This suggests the product is in active development and the team is prioritizing building over polishing the storefront. That’s a reasonable choice at this stage, though it makes it hard to evaluate the depth of the actual product from the outside.

What I find telling is the trajectory. Starting with email deliverability, finding that the TAM or competitive dynamics weren’t favorable enough, then going vertical into CRE with a phone-first approach. That’s a team that’s iterating based on market feedback rather than clinging to their original thesis.

The Verdict

The email-to-voice pivot makes strategic sense, and the CRE vertical gives them a defined wedge to build from. The founding team has the technical and business chops to execute. Whether voice cloning in sales outreach generates backlash or becomes standard practice is still an open question that affects the entire category, not just Terrakotta.

At 30 days, I’d want to see how many CRE brokerages are actively using the dialer daily and what their connect rates look like compared to manual outbound. The proof here is in usage metrics, not features.

At 60 days, the voice cloning piece needs scrutiny. Regulatory attention around AI-generated voice calls is increasing, and FCC rules on AI robocalls are getting stricter. Any company in this space needs a clear compliance story.

At 90 days, the question is whether Terrakotta stays vertical in CRE or uses it as a beachhead to expand into other high-volume outbound verticals like insurance sales, staffing, or mortgage lending. The platform play only works if the AI calling infrastructure generalizes well across different sales conversations.

The team is strong. The vertical is underserved. The timing on voice AI in sales is probably right. The risk is regulatory, and it’s real. I’d keep watching this one.